144 STELLA B. VINCENT 



Long before the rats cease to enter the cut de sacs, before any of 

 these errors are entirely cut out, the loitering at the food box 

 is no longer to be seen. It only thereafter occurs in exceptional 

 cases where an animal is entirely lost and as a consequence is 

 in a disturbed and emotional condition in which all the old errors 

 reappear. The behavior of the animals following the odor trail, 

 on the contrary, although similar at the food box was more 

 persistent than any "off trail," blind alley error. The odor, 

 it will be remembered, was that of the food with which they 

 were accustomed to be fed. Perhaps the previous stimulation 

 of the olfactory trail had made the animals more susceptible to 

 this influence. But whether, as a result of following a food odor 

 trail, all food odors attracted the attention more, or whether 

 this stronger food odor represented the natural instinctive ending 

 of a food trail and thus called a halt, these are questions for 

 thought. Either or both positions are plausible. 



Whatever the cause of this behavior, as a result of it, the 

 speed in all of the early trials was slower than that in the normal 

 maze; but by following the trail the animals were kept in the 

 true path so that the errors were greatly decreased in both the 

 initial and in the succeeding trials. 



2. The Tables 



Table 1 shows, side by side, the records for the first twenty- 

 five trials in the normal and the olfactory mazes. Figs. 1, 2, 

 and 3, show the curves plotted from these records. These curves 

 are not made like those shown in "Vision in the Maze" because 

 in the olfactory maze the learning period covered less than ten 

 trials and was practically uniform. The units used in plotting 

 were one trial, one minute and one error. Since it was the 

 following of the trail in which we were interested, the error 

 consisted in leaving the track. Returns were not counted and 

 this fact makes these curves comparable with those made for 

 the black- white maze where the returns could not be counted. 



The results of this experiment show an increase in accuracy, 

 both initial and total, over the normal maze and an increased 

 final speed. We will consider first the facts which bear out 

 these assertions as to accuracy. 



